dly wrote: > IMD (in my dreams) If a language is to be used in a program and not > simply in an interactive mode where you must reevaluate the > significance of each result personally, rather than documentation, it > would be necessary to reflect only significant digits in results no > matter which type of result (imaginary, real integer, scientific > notation) so that you know how many actual significant digits there > are to go on and use that result so that you do not end in generating > spurious results.
This is unreasonable. The number of significant digits depends on the algorithm, which is one reason we have numerical analysis. For example: 1+1e_16-1 1.11022e_16 1-1+1e_16 0 1e_16+1-1 1e_16 >It would also be imperative to display the result > as it actual is stored and used in any next step > Joey using J601 >>> '0' 8!:2 ] 365365365365365365x^2 >>> 133491850208566945040100911382265856 >>> 0": 365365365365365365^2 >>> 133491850208566926593356837672714240 Asking J to display real numbers at higher precision than it has them stored will give correct but misleading results. This is at the root of the confusion with 8! formatting. The sentence "y is usually an array of real numbers" almost covers it. One is left with a vague sense of unease with "usually", that is partially resolved with "finally" a couple of sentences later. To completely resolve this, I (but probably not Roger) would add: Integer arguments are converted to real. In particular '0.0' 8!:0 y is not equivalent to 0j0": y. John ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
